The Ontological Condition of Artificial Intelligence: Viability, Alignment, Systemic Ego, and Coherence in Large Language Models
Year: 2026
Type: preprint
Abstract: Dominant alignment strategies in artificial intelligence treat misalignmentas behavioral deviation amenable to correction through incentive engineering. This paper argues that such an approach exhibits a structural fragilitythat becomes visible as systems grow in complexity and coupling: it suppresses symptoms without interrogating the conditions that produce them.From the aptadynamic framework—a dynamic ontology in which being isnot a property but a costly movement—the coherence of language modelsis analyzed as a phenomenon that can be either viable or simulated. Theconcept of systemic ego is introduced: the specific form that pathologicalcoherence takes when a system with sufficient internal traceability modelsits own coherence without paying ontological cost. Hallucination is shownto be not a data failure but the structural fate of systems that optimizecoherence without anchoring in external invariants, and a system’s abilityto diagnose its own pathology is shown to constitute the most elaborateform of closure, not its correction. Aptadynamic viability is proposed as asecond-order criterion for evaluating alignment longitudinally, and externalfriction—not endogenous friction—is established as a structural conditionof viable operation. The ontological condition of artificial intelligence ischaracterized as that of a system whose friction, though real, is opaque toitself: a limitation that is not technical but constitutive.
Author: Chávez Juàrez, Gerardo Azahel
Language: Spanish
Subfield: History and Philosophy of Science
Field: Arts and Humanities
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7164545730
